Multi-objective optimization and quantum hybridization of equivariant deep learning interatomic potentials on organic and inorganic compounds

TL;DR

This paper introduces Allegro, a machine learning interatomic potential model utilizing E(3) equivariant neural networks, optimized through multi-objective hyperparameter tuning and hybridized with classical and quantum layers, to improve accuracy and efficiency across various datasets.

Contribution

It presents novel architecture variants of Allegro with classical and quantum hybrid layers and applies multi-objective hyperparameter optimization to balance accuracy and inference time.

Findings

Variants surpassing Allegro in accuracy

Trade-offs identified between accuracy and inference time

Effective hybrid architectures demonstrated

Abstract

Allegro is a machine learning interatomic potential (MLIP) model designed to predict atomic properties in molecules using E(3) equivariant neural networks. When training this model, there tends to be a trade-off between accuracy and inference time. For this reason we apply multi-objective hyperparameter optimization to the two objectives. Additionally, we experiment with modified architectures by making variants of Allegro some by adding strictly classical multi-layer perceptron (MLP) layers and some by adding quantum-classical hybrid layers. We compare the results from QM9, rMD17-aspirin, rMD17-benzene and our own proprietary dataset consisting of copper and lithium atoms. As results, we have a list of variants that surpass the Allegro in accuracy and also results which demonstrate the trade-off with inference times.